diff options
author | Michael Tyler <michael.tyler@arm.com> | 2022-12-15 12:39:29 +0000 |
---|---|---|
committer | michael.tyler <michael.tyler@arm.com> | 2023-01-16 09:31:00 +0000 |
commit | ba209750abc1ac7e42bab9fef5db284384d70fb3 (patch) | |
tree | 1065f242db9a9a5e48bd4a9f2fd68aef1924827a /src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp | |
parent | 8094f9dd5307c55f545b2cb41ec80a739a9b4d6f (diff) | |
download | ComputeLibrary-ba209750abc1ac7e42bab9fef5db284384d70fb3.tar.gz |
Update CPU kernels to remove x19
Resolves: COMPMID-5805
Signed-off-by: Michael Tyler <michael.tyler@arm.com>
Change-Id: I250f64531e209625e4ff176dd5a552c1c34bc484
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8909
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp')
-rw-r--r-- | src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp | 224 |
1 files changed, 112 insertions, 112 deletions
diff --git a/src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp b/src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp index 8ff0fe4dff..ce0ae29756 100644 --- a/src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp +++ b/src/core/NEON/kernels/arm_conv/depthwise/kernels/sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2022 Arm Limited. + * Copyright (c) 2022-2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -94,105 +94,105 @@ void sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst_direct_impl( "mov x3, #0x0\n" "1:" // Tile loop "str x2, [%x[params_struct], %[offsetof_args_tile_i]]\n" - "mov x21, #0x3\n" + "mov x22, #0x3\n" "str x3, [%x[params_struct], %[offsetof_args_tile_j]]\n" - "ldr x20, [%x[params_struct], %[offsetof_args_ld_input_row]]\n" - "mul x19, x2, x20\n" // offset = tile_i * ld_input_row + "ldr x21, [%x[params_struct], %[offsetof_args_ld_input_row]]\n" + "mul x20, x2, x21\n" // offset = tile_i * ld_input_row "ldr x4, [%x[params_struct], %[offsetof_args_ld_input_col]]\n" - "madd x19, x3, x4, x19\n" // offset += tile_j * ld_input_col - "mul x19, x19, x21\n" // offset *= kernel_stride * output_size + "madd x20, x3, x4, x20\n" // offset += tile_j * ld_input_col + "mul x20, x20, x22\n" // offset *= kernel_stride * output_size "ldr x5, [%x[params_struct], %[offsetof_args_inptr]]\n" - "add x5, x5, x19, LSL #2\n" // inptr[0] += offset * sizeof(float) - "add x6, x5, x20, LSL #2\n" - "add x7, x6, x20, LSL #2\n" + "add x5, x5, x20, LSL #2\n" // inptr[0] += offset * sizeof(float) + "add x6, x5, x21, LSL #2\n" + "add x7, x6, x21, LSL #2\n" "add x8, x4, x4\n" "ldr x17, [%x[params_struct], %[offsetof_args_params]]\n" - "add x16, x7, x20, LSL #2\n" + "add x16, x7, x21, LSL #2\n" "add x15, x8, x4\n" - "add x14, x16, x20, LSL #2\n" + "add x14, x16, x21, LSL #2\n" "add x13, x15, x4\n" "cbnz x3, 2f\n" - "ldr x19, [%x[params_struct], %[offsetof_args_n_tile_cols]]\n" - "sub x20, x19, x3\n" - "sub x20, x20, #0x1\n" + "ldr x20, [%x[params_struct], %[offsetof_args_n_tile_cols]]\n" + "sub x21, x20, x3\n" + "sub x21, x21, #0x1\n" "lsl x12, %x[n_channels], #0x2\n" - "mov x19, #0xc\n" - "and x20, x20, #0x3fffff\n" - "mul x19, x19, x4\n" - "orr x12, x12, x20, LSL #22\n" - "orr x12, x12, x19, LSL #38\n" - "add x25, x7, x8, LSL #2\n" - "add x24, x5, x13, LSL #2\n" - "add x23, x6, x8, LSL #2\n" - "add x22, x14, x13, LSL #2\n" - "add x21, x7, x4, LSL #2\n" - "add x20, x5, x4, LSL #2\n" - "add x19, x5, x15, LSL #2\n" - "add x11, x7, x15, LSL #2\n" - "add x10, x6, x13, LSL #2\n" - "add x9, x16, x8, LSL #2\n" - "add x28, x16, x13, LSL #2\n" - "add x27, x14, x4, LSL #2\n" - "add x26, x6, x4, LSL #2\n" - ".inst 0xf8ac4b3a // rprfm pldonce, x25, [x12]\n" - "add x25, x6, x15, LSL #2\n" - ".inst 0xf8ac48ba // rprfm pldonce, x5, [x12]\n" - ".inst 0xf8ac4b1a // rprfm pldonce, x24, [x12]\n" - "add x24, x14, x15, LSL #2\n" - ".inst 0xf8ac49da // rprfm pldonce, x14, [x12]\n" - ".inst 0xf8ac4afa // rprfm pldonce, x23, [x12]\n" - "add x23, x16, x4, LSL #2\n" - ".inst 0xf8ac4ada // rprfm pldonce, x22, [x12]\n" - "add x22, x5, x8, LSL #2\n" - ".inst 0xf8ac4aba // rprfm pldonce, x21, [x12]\n" - "add x21, x16, x15, LSL #2\n" - ".inst 0xf8ac4a9a // rprfm pldonce, x20, [x12]\n" - "add x20, x7, x13, LSL #2\n" - ".inst 0xf8ac4a7a // rprfm pldonce, x19, [x12]\n" - "add x19, x14, x8, LSL #2\n" - ".inst 0xf8ac497a // rprfm pldonce, x11, [x12]\n" - ".inst 0xf8ac48da // rprfm pldonce, x6, [x12]\n" - ".inst 0xf8ac495a // rprfm pldonce, x10, [x12]\n" - ".inst 0xf8ac4a1a // rprfm pldonce, x16, [x12]\n" - ".inst 0xf8ac493a // rprfm pldonce, x9, [x12]\n" - ".inst 0xf8ac4b9a // rprfm pldonce, x28, [x12]\n" - ".inst 0xf8ac4b7a // rprfm pldonce, x27, [x12]\n" - ".inst 0xf8ac4b5a // rprfm pldonce, x26, [x12]\n" - ".inst 0xf8ac4b3a // rprfm pldonce, x25, [x12]\n" - ".inst 0xf8ac4b1a // rprfm pldonce, x24, [x12]\n" - ".inst 0xf8ac4afa // rprfm pldonce, x23, [x12]\n" - ".inst 0xf8ac4ada // rprfm pldonce, x22, [x12]\n" - ".inst 0xf8ac4aba // rprfm pldonce, x21, [x12]\n" - ".inst 0xf8ac48fa // rprfm pldonce, x7, [x12]\n" - ".inst 0xf8ac4a9a // rprfm pldonce, x20, [x12]\n" - ".inst 0xf8ac4a7a // rprfm pldonce, x19, [x12]\n" + "mov x20, #0xc\n" + "and x21, x21, #0x3fffff\n" + "mul x20, x20, x4\n" + "orr x12, x12, x21, LSL #22\n" + "orr x12, x12, x20, LSL #38\n" + "add x27, x7, x8, LSL #2\n" + "add x26, x5, x13, LSL #2\n" + "add x25, x6, x8, LSL #2\n" + "add x24, x14, x13, LSL #2\n" + "add x23, x7, x4, LSL #2\n" + "add x22, x5, x4, LSL #2\n" + "add x21, x5, x15, LSL #2\n" + "add x20, x7, x15, LSL #2\n" + "add x11, x6, x13, LSL #2\n" + "add x10, x16, x8, LSL #2\n" + "add x9, x16, x13, LSL #2\n" + "add x28, x14, x4, LSL #2\n" + ".inst 0xf8ac4b7a // rprfm pldonce, x12, [x27]\n" + "add x27, x6, x4, LSL #2\n" + ".inst 0xf8ac48ba // rprfm pldonce, x12, [x5]\n" + ".inst 0xf8ac4b5a // rprfm pldonce, x12, [x26]\n" + "add x26, x6, x15, LSL #2\n" + ".inst 0xf8ac49da // rprfm pldonce, x12, [x14]\n" + ".inst 0xf8ac4b3a // rprfm pldonce, x12, [x25]\n" + "add x25, x14, x15, LSL #2\n" + ".inst 0xf8ac4b1a // rprfm pldonce, x12, [x24]\n" + "add x24, x16, x4, LSL #2\n" + ".inst 0xf8ac4afa // rprfm pldonce, x12, [x23]\n" + "add x23, x5, x8, LSL #2\n" + ".inst 0xf8ac4ada // rprfm pldonce, x12, [x22]\n" + "add x22, x16, x15, LSL #2\n" + ".inst 0xf8ac4aba // rprfm pldonce, x12, [x21]\n" + "add x21, x7, x13, LSL #2\n" + ".inst 0xf8ac4a9a // rprfm pldonce, x12, [x20]\n" + "add x20, x14, x8, LSL #2\n" + ".inst 0xf8ac48da // rprfm pldonce, x12, [x6]\n" + ".inst 0xf8ac497a // rprfm pldonce, x12, [x11]\n" + ".inst 0xf8ac4a1a // rprfm pldonce, x12, [x16]\n" + ".inst 0xf8ac495a // rprfm pldonce, x12, [x10]\n" + ".inst 0xf8ac493a // rprfm pldonce, x12, [x9]\n" + ".inst 0xf8ac4b9a // rprfm pldonce, x12, [x28]\n" + ".inst 0xf8ac4b7a // rprfm pldonce, x12, [x27]\n" + ".inst 0xf8ac4b5a // rprfm pldonce, x12, [x26]\n" + ".inst 0xf8ac4b3a // rprfm pldonce, x12, [x25]\n" + ".inst 0xf8ac4b1a // rprfm pldonce, x12, [x24]\n" + ".inst 0xf8ac4afa // rprfm pldonce, x12, [x23]\n" + ".inst 0xf8ac4ada // rprfm pldonce, x12, [x22]\n" + ".inst 0xf8ac48fa // rprfm pldonce, x12, [x7]\n" + ".inst 0xf8ac4aba // rprfm pldonce, x12, [x21]\n" + ".inst 0xf8ac4a9a // rprfm pldonce, x12, [x20]\n" "2:" // Tile loop: Prefetch input rows: End - "ldr x21, [%x[params_struct], %[offsetof_args_ld_output_row]]\n" - "mul x20, x2, x21\n" // offset = tile_i * ld_output_row - "mov x19, #0x3\n" + "ldr x22, [%x[params_struct], %[offsetof_args_ld_output_row]]\n" + "mul x21, x2, x22\n" // offset = tile_i * ld_output_row + "mov x20, #0x3\n" "ld1w { z18.s }, p3/Z, [x17]\n" - "ldr x26, [%x[params_struct], %[offsetof_args_ld_output_col]]\n" - "madd x20, x3, x26, x20\n" // offset += tile_j * ld_output_col - "mul x20, x20, x19\n" // offset *= output_tile_size + "ldr x27, [%x[params_struct], %[offsetof_args_ld_output_col]]\n" + "madd x21, x3, x27, x21\n" // offset += tile_j * ld_output_col + "mul x21, x21, x20\n" // offset *= output_tile_size "ld1rw { z17.s }, p3/Z, [%x[params_struct], %[offsetof_args_min]]\n" - "ldr x25, [%x[params_struct], %[offsetof_args_outptr]]\n" + "ldr x26, [%x[params_struct], %[offsetof_args_outptr]]\n" "addvl x17, x17, #1\n" - "add x25, x25, x20, LSL #2\n" // outptrs[0] += offset * sizeof(float) + "add x26, x26, x21, LSL #2\n" // outptrs[0] += offset * sizeof(float) ".inst 0xa040c220 // ld1w { z0.s-z3.s }, pn8.b/Z, [x17]\n" - "cntw x24\n" + "cntw x25\n" "addvl x17, x17, #4\n" ".inst 0xa040c224 // ld1w { z4.s-z7.s }, pn8.b/Z, [x17]\n" - "add x23, x25, x21, LSL #2\n" + "add x24, x26, x22, LSL #2\n" "whilelt p2.s, XZR, %x[n_channels]\n" "ld1rw { z16.s }, p3/Z, [%x[params_struct], %[offsetof_args_max]]\n" "addvl x17, x17, #4\n" - "cmp x24, %x[n_channels]\n" + "cmp x25, %x[n_channels]\n" "ld1w { z8.s }, p3/Z, [x17]\n" - "add x22, x23, x21, LSL #2\n" - "add x21, x26, x26\n" + "add x23, x24, x22, LSL #2\n" + "add x22, x27, x27\n" "ld1w { z9.s }, p2/Z, [x7, x8, LSL #2]\n" - "mov x20, #0x0\n" - "sub x19, XZR, x24\n" + "mov x21, #0x0\n" + "sub x20, XZR, x25\n" "ld1w { z10.s }, p2/Z, [x5]\n" "ld1w { z11.s }, p2/Z, [x5, x13, LSL #2]\n" "addvl x17, x17, #1\n" @@ -202,15 +202,15 @@ void sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst_direct_impl( "3:" // Tile loop: Channel loop "movprfx z24, z18\n fmla z24.s, p3/M, z7.s, z9.s\n" "movprfx z23, z18\n fmla z23.s, p3/M, z8.s, z9.s\n" - "whilelt p1.s, x24, %x[n_channels]\n" - "incw x20\n" + "whilelt p1.s, x25, %x[n_channels]\n" + "incw x21\n" "movprfx z25, z18\n fmla z25.s, p3/M, z6.s, z9.s\n" "fmla z24.s, p3/M, z4.s, z13.s\n" - "incw x24\n" + "incw x25\n" "mov p0.b, p2.b\n" "movprfx z26, z18\n fmla z26.s, p3/M, z5.s, z9.s\n" "movprfx z27, z18\n fmla z27.s, p3/M, z4.s, z9.s\n" - "incw x19\n" + "incw x20\n" "movprfx z28, z18\n fmla z28.s, p3/M, z3.s, z9.s\n" "fmla z23.s, p3/M, z0.s, z10.s\n" "ld1w { z10.s }, p2/Z, [x7, x15, LSL #2]\n" @@ -309,7 +309,7 @@ void sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst_direct_impl( "fmla z27.s, p3/M, z8.s, z13.s\n" "ld1w { z13.s }, p2/Z, [x14, x8, LSL #2]\n" "fmla z26.s, p3/M, z3.s, z12.s\n" - "whilelt p2.s, x20, %x[n_channels]\n" + "whilelt p2.s, x21, %x[n_channels]\n" "fmla z25.s, p3/M, z8.s, z11.s\n" "fmla z28.s, p3/M, z5.s, z11.s\n" ".inst 0xa040c220 // ld1w { z0.s-z3.s }, pn8.b/Z, [x17]\n" @@ -317,7 +317,7 @@ void sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst_direct_impl( "fmla z29.s, p3/M, z8.s, z13.s\n" "fmla z30.s, p3/M, z7.s, z13.s\n" "addvl x14, x14, #1\n" - "cmp x24, %x[n_channels]\n" + "cmp x25, %x[n_channels]\n" "fmla z31.s, p3/M, z6.s, z13.s\n" "fmax z23.s, p3/M, z23.s, z17.s\n" ".inst 0xa040c224 // ld1w { z4.s-z7.s }, pn8.b/Z, [x17]\n" @@ -327,21 +327,21 @@ void sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst_direct_impl( "ld1w { z11.s }, p1/Z, [x5, x13, LSL #2]\n" ".inst 0xc1b0ca3c // fclamp { z28.s-z31.s }, z17.s, z16.s\n" "ld1w { z12.s }, p1/Z, [x14]\n" - "st1w { z23.s }, p0, [x25]\n" + "st1w { z23.s }, p0, [x26]\n" "ld1w { z13.s }, p1/Z, [x6, x8, LSL #2]\n" - "st1w { z24.s }, p0, [x25, x26, LSL #2]\n" - "st1w { z25.s }, p0, [x25, x21, LSL #2]\n" - "addvl x25, x25, #1\n" + "st1w { z24.s }, p0, [x26, x27, LSL #2]\n" + "st1w { z25.s }, p0, [x26, x22, LSL #2]\n" + "addvl x26, x26, #1\n" "ld1w { z8.s }, p3/Z, [x17]\n" "addvl x17, x17, #1\n" - "st1w { z26.s }, p0, [x23]\n" - "st1w { z27.s }, p0, [x23, x26, LSL #2]\n" - "st1w { z28.s }, p0, [x23, x21, LSL #2]\n" + "st1w { z26.s }, p0, [x24]\n" + "st1w { z27.s }, p0, [x24, x27, LSL #2]\n" + "st1w { z28.s }, p0, [x24, x22, LSL #2]\n" + "addvl x24, x24, #1\n" + "st1w { z29.s }, p0, [x23]\n" + "st1w { z30.s }, p0, [x23, x27, LSL #2]\n" + "st1w { z31.s }, p0, [x23, x22, LSL #2]\n" "addvl x23, x23, #1\n" - "st1w { z29.s }, p0, [x22]\n" - "st1w { z30.s }, p0, [x22, x26, LSL #2]\n" - "st1w { z31.s }, p0, [x22, x21, LSL #2]\n" - "addvl x22, x22, #1\n" "blt 3b\n" "4:" // Tile loop: Channel tail "movprfx z24, z18\n fmla z24.s, p3/M, z7.s, z9.s\n" @@ -351,26 +351,26 @@ void sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst_direct_impl( "movprfx z25, z18\n fmla z25.s, p3/M, z6.s, z9.s\n" "fmla z24.s, p3/M, z4.s, z13.s\n" "ldr x2, [%x[params_struct], %[offsetof_args_tile_i]]\n" - "add x20, x2, #0x1\n" + "add x21, x2, #0x1\n" "movprfx z26, z18\n fmla z26.s, p3/M, z5.s, z9.s\n" "movprfx z27, z18\n fmla z27.s, p3/M, z4.s, z9.s\n" - "ldr x19, [%x[params_struct], %[offsetof_args_n_tile_cols]]\n" - "cmp x3, x19\n" + "ldr x20, [%x[params_struct], %[offsetof_args_n_tile_cols]]\n" + "cmp x3, x20\n" "movprfx z28, z18\n fmla z28.s, p3/M, z3.s, z9.s\n" "fmla z23.s, p3/M, z0.s, z10.s\n" "ld1w { z10.s }, p2/Z, [x7, x15, LSL #2]\n" - "ldr x19, [%x[params_struct], %[offsetof_args_n_tile_rows]]\n" + "ldr x20, [%x[params_struct], %[offsetof_args_n_tile_rows]]\n" "fmla z25.s, p3/M, z2.s, z11.s\n" "ld1w { z11.s }, p2/Z, [x7, x4, LSL #2]\n" "movprfx z29, z18\n fmla z29.s, p3/M, z2.s, z9.s\n" - "csel x2, x2, x20, LT\n" + "csel x2, x2, x21, LT\n" "fmla z24.s, p3/M, z6.s, z11.s\n" "movprfx z31, z18\n fmla z31.s, p3/M, z0.s, z9.s\n" "mov p0.b, p2.b\n" "csel x3, x3, XZR, LT\n" "fmla z23.s, p3/M, z5.s, z13.s\n" "fmla z25.s, p3/M, z3.s, z13.s\n" - "cmp x2, x19\n" + "cmp x2, x20\n" "fmla z26.s, p3/M, z2.s, z13.s\n" "fmla z27.s, p3/M, z1.s, z13.s\n" "fmla z28.s, p3/M, z0.s, z13.s\n" @@ -459,21 +459,21 @@ void sme2_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst_direct_impl( "fmax z23.s, p3/M, z23.s, z17.s\n" "fmin z23.s, p3/M, z23.s, z16.s\n" ".inst 0xc1b0ca38 // fclamp { z24.s-z27.s }, z17.s, z16.s\n" - "st1w { z23.s }, p0, [x25]\n" + "st1w { z23.s }, p0, [x26]\n" ".inst 0xc1b0ca3c // fclamp { z28.s-z31.s }, z17.s, z16.s\n" - "st1w { z24.s }, p0, [x25, x26, LSL #2]\n" - "st1w { z25.s }, p0, [x25, x21, LSL #2]\n" - "st1w { z26.s }, p0, [x23]\n" - "st1w { z27.s }, p0, [x23, x26, LSL #2]\n" - "st1w { z28.s }, p0, [x23, x21, LSL #2]\n" - "st1w { z29.s }, p0, [x22]\n" - "st1w { z30.s }, p0, [x22, x26, LSL #2]\n" - "st1w { z31.s }, p0, [x22, x21, LSL #2]\n" + "st1w { z24.s }, p0, [x26, x27, LSL #2]\n" + "st1w { z25.s }, p0, [x26, x22, LSL #2]\n" + "st1w { z26.s }, p0, [x24]\n" + "st1w { z27.s }, p0, [x24, x27, LSL #2]\n" + "st1w { z28.s }, p0, [x24, x22, LSL #2]\n" + "st1w { z29.s }, p0, [x23]\n" + "st1w { z30.s }, p0, [x23, x27, LSL #2]\n" + "st1w { z31.s }, p0, [x23, x22, LSL #2]\n" "blt 1b\n" ".inst 0xd503467f // SMSTOP\n" : : [n_channels] "r" ((unsigned long) n_channels), [offsetof_args_inptr] "I" (offsetof(Args, inptr)), [offsetof_args_ld_input_col] "I" (offsetof(Args, ld_input_col)), [offsetof_args_ld_input_row] "I" (offsetof(Args, ld_input_row)), [offsetof_args_ld_output_col] "I" (offsetof(Args, ld_output_col)), [offsetof_args_ld_output_row] "I" (offsetof(Args, ld_output_row)), [offsetof_args_max] "I" (offsetof(Args, max)), [offsetof_args_min] "I" (offsetof(Args, min)), [offsetof_args_n_tile_cols] "I" (offsetof(Args, n_tile_cols)), [offsetof_args_n_tile_rows] "I" (offsetof(Args, n_tile_rows)), [offsetof_args_outptr] "I" (offsetof(Args, outptr)), [offsetof_args_params] "I" (offsetof(Args, params)), [offsetof_args_tile_i] "I" (offsetof(Args, tile_i)), [offsetof_args_tile_j] "I" (offsetof(Args, tile_j)), [params_struct] "r" (¶ms_struct) - : "cc", "memory", "p0", "p1", "p2", "p3", "p4", "p5", "p6", "p7", "p8", "p9", "p10", "p11", "p12", "p13", "p14", "p15", "x2", "x3", "x4", "x5", "x6", "x7", "x8", "x9", "x10", "x11", "x12", "x13", "x14", "x15", "x16", "x17", "x19", "x20", "x21", "x22", "x23", "x24", "x25", "x26", "x27", "x28", "z0", "z1", "z2", "z3", "z4", "z5", "z6", "z7", "z8", "z9", "z10", "z11", "z12", "z13", "z14", "z15", "z16", "z17", "z18", "z19", "z20", "z21", "z22", "z23", "z24", "z25", "z26", "z27", "z28", "z29", "z30", "z31" + : "cc", "memory", "p0", "p1", "p2", "p3", "p4", "p5", "p6", "p7", "p8", "p9", "p10", "p11", "p12", "p13", "p14", "p15", "x2", "x3", "x4", "x5", "x6", "x7", "x8", "x9", "x10", "x11", "x12", "x13", "x14", "x15", "x16", "x17", "x20", "x21", "x22", "x23", "x24", "x25", "x26", "x27", "x28", "z0", "z1", "z2", "z3", "z4", "z5", "z6", "z7", "z8", "z9", "z10", "z11", "z12", "z13", "z14", "z15", "z16", "z17", "z18", "z19", "z20", "z21", "z22", "z23", "z24", "z25", "z26", "z27", "z28", "z29", "z30", "z31" ); } |